Non-contact Heart Rate Monitoring by Combining Convolutional Neural Network Skin Detection and Remote Photoplethysmography via a Low-Cost Camera

摘要

In this paper, we present a versatile methodology to accomplish the non-contact monitoring of heart rate signals in unconstrained environments, by combining the convolutional neural network (CNN) skin detection and the camera-based remote photoplethysmography (rPPG) methods. Compared to the widely-used three-step skin detection method (i.e., face detection, face tracking, and skin classification), the CNN method used here could enhance the monitoring robustness by achieving the skin detection in a single step. The proposed CNN-rPPG method has been tested in an unconstrained office environment to validate its applicability. Combined with the subsequent rPPG heart rate monitoring based on a low-cost camera, the method presented here is of practical interests for the large-scale deployment of the non-contact heart rate monitoring technologies.